How do I get an AI voice with the specific accent or age I actually need?

This page may contain affiliate links from which Blur Studio earns commissions.

A voice that's technically speaking the right language but the wrong accent, or that sounds twenty years off from who you're trying to reach, undercuts the message before anyone's even processed the words. This is fixable in minutes if you know which of two paths to take.

It's worth being specific with yourself about why the accent or age matters before you start, since that shapes which path is faster. If it's about basic comprehension, making sure the content is actually understandable to the audience, almost any reasonably clear voice in roughly the right accent works fine. If it's about a listener feeling genuinely spoken to, recognized as belonging to a specific group the content is for, that's a higher bar, and it's worth spending the extra time getting the match right rather than settling for "close enough."

Filter first, design second

Start by filtering an existing library by accent and approximate age. ElevenLabs' Voice Library supports both filters, and for common accents, UK, US regional, Australian, this usually turns up a workable candidate fast, with no setup cost. If nothing in the library quite lands the specific combination you're picturing, a less common accent, a very specific age-and-tone combination, Voice Design builds a new voice from a written description instead. It's available even on the free plan, so there's no cost reason to avoid trying it as a second step. For a shortlist to start from, see the best ElevenLabs voices for business.

Writing a description that actually works for accent and age

Be specific about region, not just a broad label. "British" covers a wide range; "a warm, older London accent" or "a younger, friendly Yorkshire accent" gets you closer to something usable. For age, describe energy and pacing alongside the number: "an unhurried, reassuring voice in its sixties" works better than "elderly," which can tip into a stereotype the tool renders too broadly. Generate a few rounds from slightly different wording if the first attempt doesn't land, since small changes in description tend to shift the result meaningfully.

Testing properly before you commit

1. Shortlist 2 or 3 candidates, whether filtered from the library or designed from scratch.
2. Generate each one reading your actual script, not the tool's demo or preview text, which is often short and written to flatter whatever voice reads it.
3. Listen for pacing and emphasis on your specific wording, product names, numbers, uncommon words, since these are where an accent or age mismatch actually shows up.
4. Get a second opinion from someone closer to the actual audience you're targeting than you are.

Library coverage varies a lot by accent, plan accordingly

Common accents, general American, standard British, Australian, are well covered in most libraries, with enough candidates that filtering alone usually works. Less common ones have noticeably thinner coverage. A Scottish, Welsh, South African, or Indian-English accent might turn up only a handful of library voices, sometimes none that convincingly fit a specific age or tone you're also trying to match. If you're in that situation, don't force a mediocre library match just because filtering felt like the "proper" first step; move to Voice Design sooner rather than settling for a voice that's technically in the right accent category but doesn't actually sound convincing. The time saved by skipping a frustrating library search often outweighs the extra few minutes of writing a good design prompt. For Arabic and Indian English in particular, finding a native-sounding voice goes deeper.

Avoiding an accent that reads as exaggerated or stereotyped

This is worth a direct warning, since it's an easy trap in both filtering and designing. A voice design prompt that leans on a stereotype, an exaggerated accent marker, a caricatured "old person" cadence, a regional accent played for comic effect, tends to produce something that reads as mocking or cheap to anyone from that actual background, even if that wasn't the intent at all. The same risk applies when filtering a library: the most exaggerated-sounding option in a given accent category isn't usually the most authentic one, it's often just the most performed. If you're targeting a specific regional or age-based audience, lean toward natural, understated delivery over anything that sounds like an impression of how that group talks, and have someone from that actual background listen before you commit, the same check this article already recommends for testing in general, but worth calling out specifically here.

Using different accents for different regional versions of the same content

If you're producing the same core content for multiple English-speaking markets, a US version and a UK version of the same explainer video, say, matching the accent to the audience is worth the extra setup even though the underlying script barely changes. A US audience tends to read a strong UK accent as distinctly "other," which can work in your favor or against you depending on your brand, while a UK audience often reacts the same way to a strong American accent. Rather than picking one accent and using it everywhere, filter or design a separate voice for each major market and swap it in per region, keeping everything else about the content identical. The setup cost is small (you've already done the hard part, writing the script) and it measurably improves how natural the content feels to each specific audience. To keep one voice identity across markets, see using the same voice with a different accent.

A setting worth knowing about once you're designing, not just filtering

If you end up designing a voice with a tool like ElevenLabs' rather than finding one in the library, there's usually a control for how strictly the result follows your written description versus taking creative liberty with it. Lower settings produce more variety across attempts, useful when you're still exploring what the right accent-and-age combination actually sounds like; higher settings stick closer to your exact wording but can sound slightly forced if your description was too short. Write a fuller description and explore loosely first, then tighten up once you've found a direction that's working.

What I'd tell you before you overthink this

Most people spend too long browsing and not enough time testing. A voice that sounds slightly generic but reads your real script cleanly beats a voice that sounds perfectly distinctive on a demo clip and awkward on your actual content. If you're targeting a specific regional or age-based audience for something customer-facing and ongoing, not a one-off, it's worth the extra round of testing on real material; for internal or low-stakes content, don't labor over this longer than fifteen minutes. Set yourself a hard limit before you start, fifteen minutes for low-stakes content, an hour including a second opinion for anything customer-facing and ongoing, since this is exactly the kind of task that can quietly expand to fill an afternoon if you let "just one more option" keep happening. For the brand side of the choice, see choosing an AI voice for your company videos.

A note on what actually reads as "older" or "younger" to a listener

Age in a voice comes through more in pacing and breath than in pitch alone, which is the part most people get wrong when they're picturing a result before generating anything. An unhurried, slightly longer pause between phrases reads as older far more convincingly than simply lowering a voice's pitch does, and a quicker, lighter pacing reads as younger more than raising pitch does. If a voice sounds like the right age on paper but wrong in practice, try adjusting the description toward pacing and energy rather than age as a number, since that's usually the actual lever that's missing.

A quick worked example: say you're building a voice for a product aimed at retirees learning to use a new app, and your first attempt describes "an elderly, slow voice." The result often comes out sounding frail or hesitant in a way that can feel patronizing to the actual audience, who don't think of themselves that way even if they're the target demographic. A better description, "a warm, patient voice that takes its time explaining things clearly, like a favorite teacher," gets you the pacing you actually wanted (unhurried, clear) without the baggage that comes from describing age itself as the defining trait. The lesson generalizes: describe the behavior you want (patient, clear, unhurried) rather than the demographic label (elderly), and you'll usually land on something more respectful and more usable at the same time.

When to skip this and go further

If you need the same specific accent and age combination consistently across a large volume of ongoing content, design the voice once properly, save the settings, and reuse them rather than re-picking every time. If what you actually need is a pipeline that automatically applies the right regional voice to the right content without you choosing each time, that's a short build, not a manual habit. I build that kind of thing, and if that's closer to your situation, tell me what you're producing and I'll tell you honestly whether it's worth automating.